离散变换算法在关键词检测系统中的研究

The Research on Discrete Transform Algorithm in the Keyword Spotting System

(黑龙江科技大学)【摘要】阐述了一个在较强噪声环境下针对汉语非特定说话人的连续无限制语音流中检测出其中关键词语音的公交车路线查询系统.为增强其关键词语音信号提出了建立一种新的基于离散变换的语音增强算法.并对同一噪声环境下增强关键词语音信号的模型与未增强关键词语音信号的模型进行了比较,结果显示,采用增强语音信号强度的算法在提高关键词的检测率同时,有效地降低了误报率,系统的整体性能较好,具有一定的实用性.

In this paper, the inquiry on bus route advance booking system is described, which aims at keyword speech and non - keyword speech that present to continuous unconditional speech stream for non - special speaker in Chinese small vocabulary. A new speech enhancent arithmetic based on discrete transform algorithm is offered in order to enhance the keyword speech. The comparison of keyword HMM based enhancent keyword speech with the keyword speech is done, the results show that the enhancent keyword speech has a great improvement in the probability detection for the keyword and falling the false report, the whole system has a better capability , having the practicability properly.